Charge-engineered cellulose nanofibril binders for PFAS-free, high-loading lithium battery positive electrodes

Abstract

The increasing demand for sustainable high-energy lithium-ion batteries has heightened concerns over the environmental and health impacts of per- and polyfluoroalkyl substances, used in current battery components. In particular, polyvinylidene fluoride, a synthetic resin used as the dominant electrode binder, relies on fluorinated chemicals and toxic solvents such as N-methyl-2-pyrrolidone, while also suffering from limited structural integrity under high-mass-loading conditions. Here, we show that a PFAS- and N-methyl-2-pyrrolidone-free binder based on charge-engineered cellulose nanofibrils derived from natural wood offers a renewable and sustainable alternative to polyvinylidene fluoride. The charge-engineered cellulose nanofibril binder, leveraging its cationic functional groups, promotes particle dispersion in the slurry through electrostatic repulsion and subsequently forms strong hydrogen bonds with electrode components after drying, reinforcing structural integrity. Additionally, its nanofibrous architecture forms robust, interconnected networks that facilitate electrolyte infiltration and ion transport. When implemented in LiNi_0.8Co_0.1Mn_0.1O₂ positive electrodes, this binder enables a mass loading of 113 mg cm⁻² and an electrode density of 3.65 g cm⁻³, achieving an areal capacity of 22.5 mAh cm⁻² and a volumetric energy density of 1781.5 Wh L⁻¹ at 0.05 C (corresponding to 1.13 mA cm⁻²), demonstrating competitive cell performance relative to electrodes based on conventional synthetic polymer binders. Moreover, the charge-engineered cellulose nanofibril binder enables N-methyl-2-pyrrolidone-free slurry processing, reducing the environmental footprint of electrode fabrication.

High-energy lithium-ion batteries rely on fluorinated binders that limit sustainability and performance. Here, authors develop a cellulose nanofibrils-based binder enabling PFAS-free processing while enhancing positive electrode uniformity, ion transport, and energy density in high-capacity batteries.

Introduction

Global regulatory actions are intensifying in response to the persistence, bioaccumulation, and toxicity of per- and polyfluoroalkyl substances (PFAS), a class of chemicals widely used across industrial sectors, including lithium-ion batteries^1,2. In particular, polyvinylidene fluoride (PVDF), the prevalent binder in commercial battery electrodes, raises growing concern due to its fluorinated composition³, low degradability, and reliance on a hazardous solvent N-methyl-2-pyrrolidone (NMP)^4,5. These environmental and health challenges underscore the urgent need for sustainable, PFAS-free alternatives in battery components.

In addition to its environmental drawbacks, PVDF exhibits critical technical limitations in developing high-energy-density batteries. Under high-mass-loading and high electrode density conditions, PVDF’s weak interfacial adhesion and low polarity impair structural cohesion, resulting in delamination, poor dispersion of conductive additives, and electrochemical degradation during cycling^6–11. While efforts have been made to replace PVDF with organic or bio-based binders such as synthetic copolymers and alginates^12–17, many alternatives fall short in enabling both industrially scalable manufacturing and robust electrochemical performance, largely due to limited understanding of molecular-level interactions between binders and electrode components.

Cellulose nanofibrils (CNFs), derived from abundant plant sources, offer a renewable and functional material design with high aspect ratios, tunable surface chemistry, and mechanical properties^18–22. Among their derivatives, 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO)-mediated oxidized CNF (TOCN) offers aqueous processability and strong inter-fibrillar networks. However, its densely packed hydrogen-bonding sites often lead to brittle electrode structures and hindered ion transport^23–27. Consequently, the resulting electrodes have typically relied on subtractive fabrication methods (e.g., vacuum filtration and freeze-drying) that are incompatible with slurry casting-based industrial manufacturing^28–30.

Here, we report a PFAS- and NMP-free binder strategy based on charge-engineered cellulose nanofibrils (c-CNF), designed to combine environmental safety with manufacturing scalability and high-energy electrode performance (Fig. 1a). Functionalized with quaternary ammonium groups (−N(CH₃)₃⁺), the c-CNF binder imparts electrostatic repulsion in the slurry, thereby suppressing the aggregation of electrode components and enhancing dispersion stability. After drying, it forms strong hydrogen bonds with electrode components, reinforcing interfacial adhesion and structural integrity. In addition, the nanofibrous architecture of the c-CNF forms a percolated network that provides mechanical reinforcement through physical entanglement, mitigating stress accumulation during solvent evaporation and calendering. Its open yet interconnected porous framework further ensures efficient electrolyte infiltration and facile ion transport, even at high electrode thickness and packing densities.

Fig. 1. Scalable fabrication and performance advantages of c-CNF-based high-mass-loading positive electrodes.

a Demonstration of large-scale positive electrode fabrication via the slurry-casting method using a c-CNF binder. b Schematic illustration highlighting the performance advantages of the c-CNF-based positive electrode (with a high mass loading) over a PVDF-based positive electrode, attributed to enhanced intermolecular interactions.

To validate our approach, we fabricated positive electrodes using LiNi_0.8Co_0.1Mn_0.1O₂ (NCM811) as a high-capacity electrode active material and employed ethylene glycol as a non-toxic, environmentally benign solvent. The c-CNF binder is compatible with roll-to-roll (R2R) processing, enabling the scalable production of electrodes with a high areal mass loading of 113 mg cm⁻² and a high density of 3.65 g cm⁻³ (Fig. 1b). Consequently, these electrodes achieved an areal capacity of 22.5 mAh cm⁻² and a volumetric energy density of 1781.5 Wh L⁻¹, demonstrating competitive performance relative to PVDF-based counterparts and previously reported electrodes employing conventional synthetic resin binders. By leveraging the synergistic effects of nanofibrous architecture, cationic surface engineering, and PFAS-free composition, the c-CNF binder design offers a promising route to overcome sustainability–performance trade-offs in high-energy battery electrode manufacturing.

Results

Synthesis and characterization of c-CNF binders

The c-CNF was synthesized via a two-step molecular modification process (Fig. 2a). This approach was guided by nanoscale structural engineering to achieve electrochemical stability, mechanical toughness, and dispersion uniformity. In its native state, bare CNF (b-CNF) shows fibril aggregation and poor colloidal stability. To overcome this limitation, an alkali pre-treatment (Step 1) using a sodium hydroxide (NaOH)/urea system^30–32 was employed, in which NaOH and urea act as hydrogen-bond donor and acceptor, respectively. This step disrupted inter-fibril hydrogen bonding interactions and introduced alkoxide (–O⁻ Na⁺) groups, yielding an anionic intermediate (a-CNF). Subsequent functionalization with quaternary ammonium salts (–N(CH₃)₃⁺) produced the c-CNF with tailored surface chemistry (Step 2). High-resolution transmission electron microscopy (HR-TEM) revealed the progressive fibrillation and size refinement of CNF during the functionalization (Fig. 2b–d). While the b-CNF displayed irregular aggregates, the alkali treatment partially disassembled the fibrils by weakening inter-fibril interactions. Final cationic functionalization yielded uniformly dispersed nanofibrils with an average diameter of 38 nm (Fig. 2e).

Fig. 2. Fabrication and characterization of cellulose-derived functional binders.

a Schematic diagram illustrating the fabrication process of the c-CNF binder, along with the corresponding morphological transitions at each step. HR-TEM images of b-CNF (−OH) (b), a-CNF (−O⁻ Na⁺) (c), and c-CNF (N(CH₃)₃⁺) (d). e Histograms of fiber diameter distribution. Error bars represent the mean ± standard deviation (n = 30). f–h Fourier transform infrared spectroscopy (FT–IR) spectra of different cellulose materials. i The XRD patterns of the cellulose materials. j Stress–strain curves for the films of PVDF, b-CNF, a-CNF, and c-CNF, respectively. k Photograph of the origami-folded c-CNF-based film crane.

The optimal degree of cationic substitution was determined by tuning the molar ratio of cationic agent (CA) to anhydroglucose unit (AGU) (Supplementary Fig. 1). A CA:AGU ratio of 30:1 achieved colloidal stability (zeta potential > +30 mV) and produced a semi-transparent, homogeneously dispersed c-CNF suspension (Supplementary Fig. 2). Fourier-transform infrared spectroscopy (FT–IR) demonstrated the stepwise chemical transformation of CNF. The introduction of alkoxide groups of a-CNF was evidenced by the appearance of a new C–O stretching vibration in the range of 950–1050 cm⁻¹ (Fig. 2f), while the presence of a characteristic quaternary ammonium peak at 1477 cm⁻¹ verified successful cationic functionalization of the c-CNF (Fig. 2g and Supplementary Fig. 3a). Changes in the O–H stretching band (3800–3000 cm⁻¹) further reflected modulated hydrogen bonding strength (Fig. 2h and Supplementary Fig. 3b). Compared to the a-CNF and b-CNF, the c-CNF exhibited a broadening of the hydrogen bonding region, attributed to diminished interfibrillar interactions between CNF chains induced by surface modification. Additionally, deconvolution of the hydrogen bonding region in the FT–IR spectra provided insights into fibrillar interactions through variations in peak areas (Supplementary Fig. 4). The proportion of O6-H⋯O3′ intermolecular hydrogen bonds²² (peak 1) in the b-CNF decreased from 32% to 26% upon cationic functionalization, indicating disrupted interfibrillar interactions and reduced CNF aggregation (Supplementary Fig. 5). Meanwhile, the optical transparency of the c-CNF film further verified homogeneous nanofibril dispersion, which is critical for maintaining structural homogeneity and processing stability. Scanning electron microscopy (SEM) corroborated the morphological evolution from aggregated b-CNF to the well-dispersed c-CNF structures (Supplementary Fig. 6). The oxidation stability of the c-CNF binder was assessed by linear sweep voltammetry (LSV) analysis prior to electrochemical testing. The LSV profile of the c-CNF binder film (Supplementary Fig. 7) showed a voltage stability window comparable to that of a conventional PVDF film, suggesting that the c-CNF is electrochemically stable within the operating potential window of positive electrodes under the tested conditions.

X-ray diffraction (XRD) analysis showed that the cationic functionalization also affected the crystalline microstructure of CNF. Compared to the native cellulose Iβ pattern^33–35 observed in the b-CNF, the c-CNF exhibited reduced crystallinity and transformation of the (200) to the (020) plane (Fig. 2i), suggesting lattice reorganization. This structural rearrangement contributed to mechanical toughness, with the c-CNF film achieving a toughness of 16.3 MJ m⁻³, higher than the b-CNF and a-CNF counterparts (Fig. 2j). Moreover, the c-CNF film showed high flexibility and structural robustness, retaining its structural integrity even after being subjected to repeated folding into a complex origami crane structure (Fig. 2k). Such mechanical resilience is essential for maintaining electrode integrity under the stresses induced by solvent evaporation and roll pressing during manufacturing. These results demonstrate that molecular-level engineering of CNF through targeted functionalization and structural modulation enables a binder system with optimized mechanical, chemical, and colloidal properties.

Rheological properties of c-CNF slurries

Achieving uniform dispersion and optimized rheology in electrode slurries is a prerequisite for the scalable fabrication of high-performance battery electrodes^36,37. Guided by the Derjaguin–Landau–Verwey–Overbeek (DLVO) theory^38,39, we tailored the dispersion stability and thereby modulated the rheological properties of electrode slurries. A conventional PVDF binder, which relies on weak van der Waals forces, offers limited interparticle stabilization, resulting in poor dispersion and microstructural inhomogeneity in the slurry. In contrast, the c-CNF binder introduces positive surface charges and nanofibrillar structure, thereby enhancing slurry homogeneity and colloidal stability via electrostatic repulsion (Fig. 3a). Moreover, binder-dependent differences in molecular interactions affect the shape retention of the slurry after casting. The PVDF slurry, with low shear yield stress and viscoelasticity, tends to slump and spread, whereas the c-CNF slurry maintains its cast shape due to its well-developed fibrillar network.

Fig. 3. Binder-regulated rheological properties in positive electrode slurries.

a Schematic representation of the intermolecular interactions between PVDF–PVDF and c-CNF–c-CNF chains in positive electrode slurries, respectively. b Optimized structures and molecular electrostatic potential (MEP) distributions. c Illustration of interactions between PVDF–PVDF and c-CNF–c-CNF repeating units, and d, Calculated binding energies. Cyan, pink, white, red, and blue spheres represent carbon, fluorine, hydrogen, oxygen, and nitrogen, respectively, in (b, c). e Zeta potential measurements of the binder dispersions containing conductive materials, demonstrating surface charge variations induced by functional groups. Inset, optical microscope images of the dispersion state of slurries composed of carbon black and binders. f–j Rheological properties of positive electrode slurries. Viscoelastic properties (storage modulus, G’ and loss modulus, G”) of the c-CNF binder-based positive electrode slurries with a solid content of 76 wt% as a function of shear stress (compared with PVDF binder) (f). Three-interval thixotropy test (3ITT) profiles of positive electrode slurries after structural disruption induced by an increasing shear rate (from 0.05 to 1 s⁻¹) (g). Photographs of the positive electrode slurries composed of the NCM811, carbon black, and binders (c-CNF vs. PVDF control; solid content: 76 wt%). Inset, Micro-CT images of localized agglomeration in PVDF and uniform particle distribution in c-CNF (h). Comparison of the morphological changes in positive electrode slurries (2 × 2 cm², 5 mm thick) cast after varying aging times (i). Thickness retention (t/t₀) of the positive electrode slurries (j).

Solvent evaporation during slurry drying induces vertical flow gradients that can cause component segregation, particularly in thick electrode fabrication. Evaporation at the slurry surface drives upward solvent convection, while diffusion within the slurry film acts to counterbalance the resulting concentration gradients. The relative dominance of these opposing transport mechanisms is quantified by the Peclet number (Pe = U/D₀ × h)^40,41, where U, D₀, and h denote the convection rate, diffusion coefficient, and film thickness, respectively. When Pe > 1, convection dominates over diffusion, causing upward migration of relatively lightweight binders and conductive additives, and ultimately leads to vertical inhomogeneity. This effect becomes more pronounced in thicker slurry films, as the increased film thickness (h) amplifies vertical segregation. In the PVDF slurry, weak van der Waals interactions and low viscosity produce minimal structural resistance to the flow, resulting in Pe ≫ 1 and severe segregation (Fig. 3a, left). In contrast, after slurry casting, the c-CNF binder forms a percolated and highly interconnected nanofiber network via hydrogen bonding between CNFs. This network imparts high viscosity and shear yield stress, which inhibit solvent flow and suppress phase migration, thereby maintaining the dispersion stability throughout the electrode film. This suppression of convective flow corresponds to a reduction in the effective Peclet number, minimizing vertical component migration during drying (Fig. 3a, right).

This behavior is further described by Darcy’s law (U = –k/μ ∇P)⁴², where k, μ, and ∇P represent the transport permeability, dynamic viscosity, and pressure gradient, respectively. The high entanglement density of the c-CNF network increases the viscous resistance (μ), thereby reinforcing the flow-suppressing mechanism already captured by the Peclet analysis. As a result, the c-CNF electrode exhibits uniform composition across the electrode thickness and is free from surface cracking, even at high mass loadings. This structural homogeneity is critical for achieving scalable, defect-free electrode production and reliable electrochemical performance.

Density functional theory (DFT) calculations were carried out to elucidate the molecular basis of the binder interactions in the slurry. Molecular electrostatic potential (MEP) mapping revealed a pronounced positive charge distribution around the quaternary ammonium groups in the c-CNF (Fig. 3b), indicating stronger electrostatic interactions than those in the PVDF binder (Supplementary Data 1 and 2). In addition, hydrogen bonding between –OH groups of the c-CNF exhibited a binding energy of –0.64 eV, almost an order of magnitude higher than the –F···H– interactions in PVDF (–0.07 eV) (Fig. 3c, d).

Surface zeta potential measurements verified the enhanced electrostatic stabilization enabled by the c-CNF. Surface charge is a key determinant of slurry stability, influencing interparticle interactions and aggregation resistance^38,43. To identify the role of the binders, model suspensions without electrode active materials were prepared. While PVDF, b-CNF, and a-CNF showed low negative potentials (–8, –12, and –21 mV, respectively), the c-CNF slurry exhibited a high positive zeta potential of +31.9 mV (Fig. 3e and Supplementary Fig. 8). Visual stability tests using the model suspensions confirmed the enhanced dispersion in the c-CNF system (Supplementary Fig. 9), highlighting the function of electrostatic interactions in regulating slurry microstructure.

The c-CNF slurry was prepared by redispersing freeze-dried c-CNF in ethylene glycol, a non-toxic and environmentally benign solvent, followed by the sequential addition of carbon black additives and NCM811 particles (Supplementary Fig. 10). The slurries were formulated with a fixed solid content of 76 wt% and compositions of 97:2:1 and 96:2:2 (NCM811/carbon black additive/binder, w/w/w). Viscosity profiles revealed shear-thinning behavior^44,45 in both slurries, but the c-CNF slurry exhibited pronounced non-Newtonian characteristics (Supplementary Fig. 11) that enabled adaptive viscosity under shear and more uniform stress distribution during processing. Oscillatory shear stress sweep analysis further demonstrated the thixotropic nature of the c-CNF slurry, with the crossover point between storage (G’) and loss (G”) moduli, an indicator of shear yield stress (τ_y), occurring at a significantly higher shear yield stress (120 Pa) compared to the PVDF slurry (0.2 Pa), reflecting a well-developed viscoelastic structuring (Fig. 3f).

This rheological strength translated into more stable coating behavior. The maximum achievable slurry thickness (h_max)⁴⁶ was estimated to be 2.73 mm for the c-CNF, almost double that of the PVDF (1.35 mm) (Supplementary Fig. 12). Achieving structural recovery and dispersion uniformity in electrode slurries is essential for scalable, high-mass-loading electrode fabrication^47,48. Three-interval thixotropy tests (3ITTs), which measure viscosity recovery under increasing shear rates^38,45, showed almost complete viscosity recovery in the c-CNF slurry, in contrast to poor recovery in the PVDF counterpart (Fig. 3g). This behavior stems from the synergistic interplay between the percolated nanofiber network and hydrogen bonding, which enhances slurry resilience under dynamic shear deformation.

Micro-computed tomography (micro-CT) images of the slurries further demonstrated the structural differences: the PVDF slurry showed localized agglomeration, while the c-CNF slurry exhibited uniform microstructures with continuous particle distribution (Fig. 3h). Casting tests using 2 × 2 cm² molds showed that the c-CNF slurry retained its shape with minimal deformation, maintaining ~92% of its original thickness after 1 min. In contrast, the PVDF slurry collapsed, retaining only ~50% of its original thickness (Fig. 3i, j). These results underscore the importance of molecularly engineered binders, wherein the electrostatic stabilization improves dispersion homogeneity, and the hydrogen bonding regulates viscoelasticity, together ensuring the consistency and processability essential for scalable electrode fabrication.

Structural homogeneity and dimensional integrity of c-CNF electrodes

The effect of binders on electrode structure was investigated, with a focus on their intermolecular interactions with electrode components. The PVDF binder, characterized by weak van der Waals interactions, exhibited limited affinity for electrode active materials and current collectors, resulting in a heterogeneous distribution of electrode components and structural instability (Fig. 4a). In contrast, the c-CNF binder forms strong hydrogen-bonding interactions with both NCM811 particles and current collectors, strengthening the structural integrity of the resulting electrode (Fig. 4b).

Fig. 4. Through-thickness uniformity and structural stability of c-CNF electrodes.

a, b Proposed mechanism of PVDF and c-CNF binders in positive electrodes, based on intermolecular interactions. c–h Structural analysis of the c-CNF electrode (compared with PVDF electrode), in which the positive electrode thickness was set as 120 μm (areal mass loading of 40 mg cm⁻²). Cross-sectional backscattered scanning electron microscopy (BS-SEM) of the PVDF electrode (c). Field emission-electron probe microanalyser (FE-EPMA) mapping images of cross-sections of PVDF electrode (d). Magnified cross-sectional SEM images of PVDF electrode (e). BS-SEM of the c-CNF electrode (f). FE-EPMA mapping images of cross-sections of c-CNF electrode (g). Magnified cross-sectional SEM images of c-CNF electrode (h). i Quantitative comparison of work of adhesion between PVDF and c-CNF binders with NCM811 and the Al current collector. j XPS spectra of Ni 2p for PVDF and c-CNF binder-based NCM811 positive electrodes at varying etching depths. k Properties of electrodes; adhesion strength (N m⁻¹), electrical conductivity (S cm⁻¹), and specific surface area per unit volume (m⁻¹). l Wettability of electrodes; contact angle between electrodes and liquid electrolyte (θ), ionic tortuosity, and MacMullin number.

Backscattered scanning electron microscopy (BS-SEM) and field-emission electron probe microanalyser (FE-EPMA) revealed significant differences in the internal morphology. The PVDF electrode showed an uneven distribution of conductive additive and binder, accompanied by agglomeration (Fig. 4c–e and Supplementary Fig. 13). To quantitatively assess the chemical affinity and intermolecular interactions between electrode components, we calculated the work of adhesion values using the Fowkes equation^49,50. The contact angle measurements were conducted using water and diiodomethane as probe liquids to determine surface energy. For the PVDF, the binder-binder interaction (work of cohesion = 59.58 mN m⁻¹, Supplementary Fig. 14a) was lower than its interactions with active materials (78.40 mN m⁻¹) and conductive additives (77.56 mN m⁻¹) (Supplementary Fig. 15 and Supplementary Table 1). This imbalance resulted in binder clustering, phase segregation of electrode components, and formation of dense morphologies. In contrast, the c-CNF binder exhibited well-balanced intermolecular interactions with electrode active materials (107.1 mN m⁻¹) and conductive additives (103.6 mN m⁻¹) (Supplementary Fig. 15 and Supplementary Table 1) comparable to binder-binder interactions (106.9 mN m⁻¹) (Supplementary Fig. 14b). This balanced intermolecular interaction, combined with the synergistic effect of uniformly high viscoelasticity in the electrode slurry (as shown in Fig. 3), enhances structural stability during slurry casting, thereby preventing gravitational deformation and mitigating stress accumulation upon solvent drying. As a result, the electrode components are homogeneously dispersed throughout the electrode (Fig. 4f–h), ensuring a consistent microstructure essential for high-mass-loading electrode fabrication.

Furthermore, to evaluate the manufacturing processability of electrodes, laser scanning confocal microscopy (LSCM)⁵¹ was used to measure height variations of dried electrodes. The PVDF electrode showed remarkable surface non-uniformity, which induces uneven stress formation during the calendering process (Supplementary Fig. 16a). By comparison, the c-CNF electrode exhibited minimal height variation, demonstrating the enhanced calendering tolerance and structural homogeneity (Supplementary Fig. 16b).

The c-CNF binder exhibited higher adhesion strengths to both NCM811 (107.1 mN m⁻¹) and the Al current collector (88.53 mN m⁻¹) compared to PVDF (78.40 mN m⁻¹ and 68.77 mN m⁻¹, respectively) (Fig. 4i and Supplementary Table 1). This result was validated by X-ray photoelectron spectroscopy (XPS) depth profiling (Fig. 4j and Supplementary Fig. 17), where the c-CNF electrode exhibited a systematic shift of the Ni peak to lower binding energies, decreasing from 854.9 eV at the surface to 853.9 eV in the bulk with increasing etching depth. This shift suggests that the cationic functional groups in the c-CNF binder interact with oxygen atoms in the Ni–O lattice^52,53 of NCM811, modulating the local electronic environment around Ni. In contrast, the Ni peak in the PVDF electrode remained nearly unchanged (~854.1 eV), suggesting weak interfacial interactions. Molecular simulations (MD) further supported this interpretation. The c-CNF binder (N(CH₃)₃⁺∙∙∙Al₂O₃) exhibited a binding energy of –1193.8 kJ mol⁻¹, substantially stronger than that of the PVDF binder (F∙∙∙Al₂O₃, –530.26 kJ mol⁻¹) (Supplementary Fig. 18 and Supplementary Data 3–9). This trend was further supported by electrode adhesion measurements using peel-off tests and the Surface and Interfacial Cutting Analysis System (SAICAS) (Fig. 4k, left, and Supplementary Fig. 19–21). The PVDF electrode showed detachment from the current collector, indicating weak adhesion and structural instability. In contrast, the c-CNF electrode remained intact, demonstrating stronger interfacial adhesion and mechanical integrity, particularly in high areal mass loading applications. We further assessed electrode mechanical integrity under electrolyte-immersed conditions (Supplementary Fig. 22). Compared with PVDF, the c-CNF electrode better preserved its architecture and suppressed immersion-induced mechanical degradation and interfacial detachment, indicating robust stability after electrolyte exposure.

This higher structural homogeneity also translated into superior electrical conductivity. The electrical conductivity of the c-CNF electrode reached 0.23 S cm⁻¹, higher than that of the PVDF electrode (0.16 S cm⁻¹) due to its continuous electron conduction network (Fig. 4k, middle). In addition, the fibrous structure of the c-CNF contributed to the formation of open porous structures conducive to electrolyte transport. This structural feature of the c-CNF electrode was quantified by analyzing the electrochemically active specific surface area per unit volume (i.e., the effective surface area of electrode active materials in contact with liquid electrolytes) using genetic algorithm modelling^54,55. The c-CNF electrode (16.19 × 10⁵ m⁻¹) exhibited a higher specific surface area compared to the PVDF electrode (0.2591 × 10⁵ m⁻¹) (Fig. 4k, right), indicating that the c-CNF maintains open and interconnected networks, whereas PVDF tends to aggregate upon solvent drying, forming a denser structure that limits electrolyte accessibility.

The c-CNF electrode also exhibited enhanced electrolyte wettability, as evidenced by lower contact angles (Fig. 4l, left, and Supplementary Fig. 23). Although the porosity levels were similar (30.9% for the c-CNF electrode and 29.7% for the PVDF counterpart), the enhanced electrolyte wettability of the c-CNF electrode is ascribed to its cationic surface chemistry. This promotes favorable electrostatic interactions with the liquid electrolyte, which extends beyond differences in structural porosity alone. This enhanced electrolyte affinity was directly translated into reduced ionic tortuosity (Fig. 4l, middle, Supplementary Fig. 24) and lower MacMullin number (Fig. 4l, right, Supplementary Fig. 25), enabling faster ion diffusion and greater penetration depth across varying current densities (Supplementary Fig. 26). This stable interface and better electrolyte wettability of the c-CNF binder were further corroborated by galvanostatic intermittent titration technique (GITT) analysis. The c-CNF electrode showed consistently lower internal resistance throughout the charge–discharge cycles (Supplementary Fig. 27), validating the ability of the binder to minimize polarization and maintain stable electrochemical interfaces.

Electrostatic regulation of ion transport by c-CNF binders

Efficient ion transport within electrodes is critical for achieving uniform reaction kinetics and high energy density in batteries^51,56. The cationic quaternary ammonium groups (−N(CH₃)₃⁺) of the c-CNF binder play a vital role in modulating ion transport by electrostatically interacting with PF₆⁻ in the liquid electrolyte, thereby accelerating Li⁺ mobility in the electrodes (Fig. 5a). To elucidate these interactions, molecular simulations were conducted using quaternary ammonium–anion pairs (Cl⁻, PF₆⁻, and TFSI⁻) confined within the c-CNF (Supplementary Fig. 28a and Supplementary Data 10–12). The MEPs of the anions were compared (Supplementary Fig. 28b), revealing that TFSI⁻ exhibited the lowest mean MEP (−0.025 Ha e⁻¹) among the tested anions. The large molecular volume of TFSI⁻ (1268 bohr³ mol⁻¹) allows its low electrostatic potential to spread uniformly over the entire structure (Supplementary Fig. 28c), facilitating its detachment from the c-CNF and promoting efficient ion exchange with PF₆⁻ in the liquid electrolyte. The thermodynamics of ion exchange between c-CNF and LiPF₆ salt in the electrolyte was assessed by computing the free energy difference between c-CNF paired with different anions and the Li⁺–PF₆⁻ complex. Among the tested anions, TFSI⁻ exhibited the lowest free energy change (Δ(ΔG⁰) = −46.7 kJ mol⁻¹), indicating that its replacement by PF₆⁻ from LiPF₆ occurs spontaneously (Supplementary Fig. 28d).

Fig. 5. Enhancement of ionic transport within positive electrodes enabled by the c-CNF binder.

a Schematic illustration showing the role of the c-CNF binder in enabling the enhanced electrolyte wettability through electrostatic attraction with electrolyte anions. b FT–IR spectra (focusing on P–F vibration) of the electrolyte, electrolyte with PVDF binder, and electrolyte with c-CNF binder, respectively. c Radial distribution function (RDF) profiles of the P atom in PF₆⁻ at the binder-electrolyte interface. d ⁷Li MAS NMR spectra of PVDF and the c-CNF binder. e Nyquist plots obtained from a symmetric cell at 0% state of charge (SOC), where symbols and solid lines represent the experimental data and fitted curves, respectively. f Comparison of electrochemical impedance characteristics of PVDF– and c-CNF–based positive electrodes (mass loading of 40 mg cm⁻²) in a half-cell configuration using DRT analysis. g–i Partial differential equations (PDE)-based simulation. Schematic representation of the PDE-based modeling and simulation process (g). Lithium-ion concentration profile along the electrode thickness (h). Probability density of the fitted apparent Li⁺ diffusion coefficients (i).

FT–IR analysis (Fig. 5b) of the c-CNF electrode revealed a shift of the P–F bond vibration peak to lower wavenumber, from 840 to 836 cm⁻¹, indicating strong electrostatic interactions between the positively charged quaternary ammonium moieties in the c-CNF and PF₆⁻ in the liquid electrolyte. To elucidate the ion distribution at the binder–electrolyte interface, the radial distribution function (RDF) of PF₆⁻ was analyzed (Fig. 5c and Supplementary Data 13–16). The RDF profile of PF₆⁻ for the c-CNF binder exhibited a distinct peak at r = 4.1 Å, accompanied by a relatively high coordination number (CN), indicating strong electrostatic attraction between the c-CNF binder and PF₆⁻. In contrast, the PVDF binder showed a peak at r = 4.3 Å with a lower CN, suggesting weaker interactions with PF₆⁻ and less effective ion coordination within the electrolyte-filled electrodes. This result was further supported by the measured Li⁺ transference number ($t_{{\mathrm{Li}}^{+}}$) of binder films swollen with a 1 M LiPF₆ electrolyte in ethylene carbonate (EC)/ethyl methyl carbonate (EMC) (3:7, v/v), used as a model system. The c-CNF film exhibited a higher $t_{{\mathrm{Li}}^{+}}$ value of 0.83 compared to 0.54 for the PVDF film with identical liquid electrolyte conditions (Supplementary Fig. 29 and Supplementary Table 2). These results underscore the critical role of c-CNF’s cationic functional groups in modulating electrolyte interactions and facilitating Li⁺ transport within the electrode. To provide additional insight into the local chemical environment of Li⁺ ions, ⁷Li magic-angle spinning (MAS) nuclear magnetic resonance (NMR) spectroscopy was employed to analyze positive electrodes fabricated with different binders (Fig. 5d). The ⁷Li MAS NMR spectrum of the c-CNF-based positive electrode exhibited a singlet resonance at −0.338 ppm, which is downfield-shifted relative to the corresponding signal observed for the PVDF–based electrode (−0.471 ppm). This downfield shift indicates facilitated dissociation of lithium salts in the presence of the c-CNF binder, reflecting an altered Li⁺ coordination environment^51,56. Such an effect is expected to be favorable for Li⁺ transport within the electrode, thereby contributing to the competitive electrochemical performance of the c-CNF-based electrodes.

Electrochemical impedance spectroscopy (EIS) was conducted using a symmetric blocking cell configuration⁵⁷ to evaluate electrolyte wettability and ion transport resistance (R_ion) within electrolyte-filled electrodes with varying areal mass loadings (Supplementary Fig. 30). The equivalent circuit elements, fitted resistance values, and fitting errors, including χ² and χ²/|Z|², are summarized in Supplementary Table 3. At a moderate areal mass loading of 20 mg cm⁻², both electrodes exhibited comparable EIS profiles and R_ion values, indicating similar initial ion transport characteristics. However, as the loading increased to 30 mg cm⁻², the c-CNF electrode exhibited a slightly lower R_ion compared to the PVDF electrode. This difference became more pronounced at a higher areal mass loading of 40 mg cm⁻² (Fig. 5e), reflecting facilitated ion conduction in the high-mass-loading electrodes.

To further elucidate how the reduced ion-transport resistance influences the overall electrode impedance, distribution of relaxation times (DRT) analysis was performed on the EIS spectra. The DRT-EIS results reveal that, compared to the PVDF–based electrode, the c-CNF-based electrode exhibits lower contributions from both charge-transfer resistance and diffusion-related polarization (Fig. 5f). These reduced resistive components indicate more efficient interfacial charge-transfer kinetics and facilitated ion transport within the electrode, particularly under high areal mass loading conditions.

To gain deeper insight into the electrochemical behavior of the c-CNF electrode described above, a partial differential equation (PDE)-based electrochemical model was developed to simulate its performance under practical operating conditions. This model incorporated experimentally measured parameters, including electrode tortuosity, open-circuit voltage, resistance, and other relevant properties. Key model parameters related to the quantitative electrochemical performance of materials were identified by fitting the simulation outputs to experimental data, enabling accurate parameter estimation. To account for uncertainty in the optimization process, parameter estimation was repeated 100 times using genetic algorithms. The resulting parameter distributions were converted into probability density distributions by normalizing each distribution such that the integrated area equals one; accordingly, the y-axis in Fig. 5i is labeled as “Probability density”^58,59. These calibrated parameters were then utilized to predict the electrochemical behavior of the cells under various operating conditions (Fig. 5g and Supplementary Fig. 31). The distribution of Li⁺ concentration across the electrode thickness was simulated based on the electrolyte–binder interaction and wettability characteristics. Regardless of the state of charge (SOC), the c-CNF electrode consistently exhibited higher Li⁺ concentrations throughout the entire electrode depth compared to the PVDF electrode. This can be attributed to the facilitated electrolyte accessibility and ion transport enabled by the favorable interaction between the c-CNF binder and the electrolyte (Fig. 5h). This enhanced ionic accessibility was reflected in the fitted apparent Li⁺ diffusion coefficients shown in Fig. 5i. The higher most probable values of the apparent Li⁺ diffusion coefficients for the c-CNF electrode, compared with the PVDF electrode, suggest more effective local Li⁺ transport and active material utilization.

Enhanced electrochemical performance of c-CNF electrodes

The electrochemical performance was evaluated using NCM811 electrodes with a thickness of ~120 μm and an areal mass loading of 40 mg cm⁻², paired with Li-metal (thickness = 100 μm). The c-CNF electrode delivered an areal capacity of 8 mAh cm⁻² and a specific capacity of 200 mAh g_NCM811⁻¹ at a current density of 0.1 C (= 0.8 mA cm⁻²), which almost reached the theoretical capacity^51,60 of NCM811 (Supplementary Fig. 32). In comparison, the PVDF electrode showed lower capacities due to insufficient ion/electron pathways throughout the electrode. The rate capability of the PVDF and c-CNF electrodes was examined by varying charge/discharge current densities from 0.2 C (= 1.6 mA cm⁻²) to 1.0 C (Fig. 6a and Supplementary Fig. 33). The difference in specific capacity between the PVDF and c-CNF electrodes became more pronounced at higher current densities and areal mass loadings. This result underscores the critical role of the c-CNF binder in enhancing ion transport efficiency and reducing polarization losses, ultimately improving the rate capability of the cells.

Fig. 6. Electrochemical evaluation of c-CNF-based positive electrodes.

a Rate capability evaluated at various current densities, where 1 C corresponds to 8 mA cm⁻² for electrodes with an areal mass loading of 40 mg cm⁻². b Cycling performance at a charge/discharge current density of 0.33 C/0.33 C (=2.67 mA cm⁻²) within a voltage range of 3.0–4.3 V with an areal mass loading of 40 mg cm⁻² and an electrode density of 3.35 g cm⁻³. Data are presented as mean values from three independently tested cells (n = 3). c, d Cross-sectional SEM images and depth-dependent Raman spectra with corresponding E_g/A_1g intensity ratios collected from the top and bottom regions of PVDF- and c-CNF-based electrodes with a mass loading of 40 mg cm⁻². For c, d, cells were cycled at 0.5 C for 5 cycles at 23 ± 1 °C and disassembled in the fully discharged state (corresponding to 0% state of charge), delivering an areal capacity of ~8 mAh cm⁻². e ToF-SIMS mapping images of the byproducts formed on the surface of the electrodes. f Cross-sectional SEM images of the cycled NCM811 in PVDF and c-CNF electrodes. For e, f cells were cycled at 0.33 C for 80 cycles at 23 ± 1 °C and disassembled in the fully discharged state, with a discharge areal capacity of ~8 mAh cm⁻². g Galvanostatic charge/discharge profiles of the pouch cells (c-CNF / Li-metal) at a current density of 0.1 C/0.1 C within a voltage range of 3.0–4.3 V and areal mass loading of 40 mg cm⁻².

In addition, the c-CNF electrode exhibited enhanced capacity retention compared to the PVDF electrode (Fig. 6b and Supplementary Fig. 34). Meanwhile, after replacing the cycled Li-metal, the c-CNF electrode successfully returned to its capacity, whereas the PVDF electrode failed to recover. This result indicates that the capacity degradation of the c-CNF electrode is attributed to Li-metal deterioration. To elucidate this cyclability difference between the c-CNF and PVDF electrodes, a postmortem analysis was performed on the cycled electrodes.

Confocal Raman mapping was conducted on cycled electrodes at a fully discharged state (depth of discharge, DOD = 100%) to probe redox heterogeneity across the electrode thickness^51,61,62. The intensity ratio of the E_g/A_1g vibrational modes, which correlates with the degree of lithiation in Ni-rich layered oxide cathodes, was used as a quantitative indicator of the local redox state. As shown in Fig. 6c, the PVDF–based cathode exhibited a pronounced variation in the E_g/A_1g ratio across the electrode thickness, indicating surface-localized lithiation and incomplete utilization of the electrode depth. This redox inhomogeneity is consistent with the non-uniform component dispersion (Fig. 4a–h) and the resulting Li⁺ concentration gradient, which limits effective ion transport to deeper regions of the electrode. In contrast, the c-CNF-based electrode showed a nearly constant E_g/A_1g ratio throughout the entire electrode thickness (Fig. 6d), demonstrating a homogeneous DOD distribution and uniform redox activity. Such redox uniformity mitigates localized current-density accumulation and suppresses non-uniform volume changes of the active material particles during cycling^63,64.

The F 1s XPS spectra (Supplementary Fig. 35) revealed that the c-CNF electrode exhibited lower peak intensities for cathode–electrolyte interphase (CEI)-related species compared to the PVDF electrode, including the Li_xPO_yF_z peak at 686.7 eV, a known marker of CEI byproducts^65,66. This reduction in CEI byproducts demonstrates the role of the c-CNF binder in suppressing unwanted side reactions, thereby mitigating electrode material degradation and improving interfacial stability. Time-of-flight secondary ion mass spectrometry (ToF-SIMS) mapping (Fig. 6e) further confirmed the suppression of byproduct formation in the c-CNF cathode compared to the PVDF electrode.

This result was verified by analyzing the structural changes of the cycled NCM811. The cycled NCM811 in the PVDF electrode showed severe cracking (Fig. 6f), consistent with non-uniform redox reactions and localized current-density variations that can arise from the inhomogeneous distribution of NCM811, carbon black additives, and binder. Crystallographic analysis revealed substantial structural degradation in both the surface and bulk regions of the PVDF electrode, with a loss of crystallinity (Supplementary Fig. 36a, b). Such degradation is particularly pronounced in high-mass-loading electrodes because the areal current density increases with areal capacity (i = C-rate × areal capacity), imposing a higher current per unit area than in low-loading electrodes. Under these conditions, non-uniform PVDF distribution can amplify local ionic-transport and electronic-percolation limitations, elevating local overpotentials and reaction-rate heterogeneity. This heterogeneity accelerates localized (de)lithiation and interfacial side reactions, generating strong SOC/redox gradients and redox-induced chemo-mechanical mismatch that concentrates stress at grain boundaries and interparticle interfaces, thereby promoting crack initiation and propagation and particle fragmentation. In contrast, the c-CNF electrode showed suppressed crack formation and greater structural integrity, retaining crystallinity after cycling (Supplementary Fig. 36c, d). These results highlight the viability of the c-CNF binder in mitigating mechanical degradation while maintaining the crystallinity of the electrode active material, contributing to prolonged cycle life and structural durability of the electrodes.

Pouch cells (area = 2 × 2 cm²) containing the c-CNF positive electrode (8 mAh cm⁻²) and a Li-metal negative electrode (thickness = 100 μm, N (negative)/P (positive) capacity ratio = 2.5) were fabricated to validate the scalability of this approach. The PVDF electrode showed a significant capacity loss in the pouch cell configuration. This result demonstrates the inherent limitations of the PVDF binder, where non-uniform dispersion of the electrode components and weak adhesion to the current collector hinder charge transport, leading to capacity loss and reduced electrochemical stability. In contrast, the c-CNF electrode approached the theoretical capacity of NCM811 even in the pouch cell (Fig. 6g). This finding underscores the viability of the c-CNF binder in maintaining uniform electrode microstructures.

High-mass-loading and high-density electrodes enabled by c-CNF binders

Achieving both high areal mass loading and high electrode density remains a critical challenge in the development of high-density batteries, as it requires simultaneous optimization of electronic/ionic conductivity, structural integrity, and process scalability. Single-walled carbon nanotubes (SWCNTs) were introduced as a high-aspect-ratio conductive additive owing to their established role in enhancing electrical conductivity^30,67,68. However, dispersion remains a challenging hurdle, as SWCNT aggregation may impair electronic conduction pathways, increase charge transfer resistance, and degrade electrochemical performance. In this work, the c-CNF binder facilitated uniform SWCNT dispersion via noncovalent interactions, including hydrophobic interactions with relatively nonpolar domains (Supplementary Fig. 37a). Raman spectroscopy further supported these interactions, revealing characteristic π–π stacking features that contribute to enhanced dispersion stability of SWCNTs (Supplementary Fig. 37b). Consequently, well-distributed electronic networks were formed in the electrode structure (Supplementary Fig. 37c). In addition, the effect of SWCNT incorporation on slurry processability was examined by analyzing the rheological behavior of the electrode slurry. The electrode slurries composed of NCM811, carbon black additives, and binders exhibited high viscoelasticity (Supplementary Fig. 38), comparable to those of carbon black additives-based slurries observed in Fig. 3f, g.

The electrode formulation was adjusted with a composition ratio of 98.5:0.5:1 (active material/conductive additive/binder, w/w/w). Despite this low binder content, c-CNF electrodes with areal mass loadings of 80, 90, and 113 mg cm⁻² were successfully fabricated, while maintaining a high electrode density of 3.65 g cm⁻³ (Fig. 7a). Cross-sectional morphologies of the c-CNF electrodes were provided as a function of their areal mass loading (Supplementary Fig. 39), demonstrating the uniform distribution of electrode components and structural integrity across all mass loadings.

Fig. 7. High-mass-loading and electrode-density c-CNF electrodes.

a Thickness and areal capacity of the c-CNF electrodes as a function of areal mass loading after roll pressing. Symbols represent experimental data, and lines represent the theoretical thickness calculated from the areal mass loading with a constant electrode density (ρ) of 3.65 g cm⁻³. Inset, photographs (cross-sectional view) of c-CNF electrodes as a function of areal mass loading. b Photographs of PVDF-based electrodes (M/A = 50 mg cm⁻², ρ = 3.35 g cm⁻³); c photographs of c-CNF-based electrodes (M/A = 90 mg cm⁻², ρ = 3.65 g cm⁻³). d Specific capacity of the c-CNF positive electrode as a function of the areal mass loading at a current rate of 0.05 C (1 C = 200 mA g⁻¹) and voltage range of 3.0–4.3 V. The dotted line represents the theoretical capacity of the NCM811 (=200 mAh g⁻¹). Inset, SEM images (cross-sectional view) of the c-CNF electrodes as a function of areal mass loading. e Galvanostatic charge/discharge profiles of the cells with different positive electrode areal mass loadings, plotted as a function of areal capacity, at a current rate of 0.05 C (1 C = 200 mA g⁻¹). f Cell energy density of various electrodes as a function of areal capacity of c-CNF electrode, compared with previously reported cathode binders (described in Supplementary Table 6). g Top-view photographs of PVDF– and c-CNF–based pouch cells with identical total cell capacity. Bottom images show the difference in thickness between the two cells. h Comparison of the weight of pouch cells fabricated with PVDF– and c-CNF–based electrodes. i Cell cost and electrode-manufacturing energy demand as a function of areal capacity, estimated using the BatPaC model^41,62 for a 72 Ah lithium-ion cell with a graphite||NCM811 configuration. j A comprehensive comparison of the major characteristics between the c-CNF binder in this work and PVDF. The criteria of the green score of binder and solvent are described in Supplementary Table 8.

Whereas the PVDF-based electrodes showed cracking and detachment under an areal mass loading of 50 mg cm⁻² and an electrode density of 3.35 g cm⁻³ (Fig. 7b), the c-CNF electrode maintained its structural integrity even at simultaneously high areal mass loading (90 mg cm⁻²) and electrode density (3.65 g cm⁻³) (Fig. 7c). Notably, the c-CNF electrodes maintained their capacities close to the theoretical specific capacity of NCM811 across all tested mass loadings (Fig. 7d), exhibiting linear areal capacity scaling up to 22.5 mAh cm⁻² at 113 mg cm⁻² (Fig. 7e). This result indicates the efficient utilization of electrode active materials even at such high mass loadings. Moreover, the c-CNF electrodes exhibited stable cycling performance across various mass loading conditions (Supplementary Fig. 40).

The simultaneous achievement of both high areal mass loading and high electrode density has rarely been reported in previous studies, underscoring the promising potential of the c-CNF binder in advancing high-energy batteries. This unique capability enables electrodes to meet the demands for both high areal capacity and high energy density, which are key performance metrics for practical energy storage applications. For example, at an areal mass loading of 113 mg cm⁻², the c-CNF electrode achieved an areal capacity of 22.5 mAh cm⁻² (a gravimetric energy density of 431.8 Wh kg⁻¹, calculation details in Supplementary Table 4) and a volumetric energy density of 1781.5 Wh L⁻¹, based on experimentally measured parameters (calculation details in Supplementary Table 5), validating its capability to sustain high energy output while maintaining structural integrity. Remarkably, these metrics were achieved with only 1 wt% binder, while still maintaining a high electrode density of 3.65 g cm⁻³.

The dashed lines represent the calculated cell energy density (Wh L⁻¹) as a function of areal capacity at different electrode densities (3.00, 3.35, and 3.65 g cm⁻³) (Fig. 7f and Supplementary Table 6). The cell energy density increases with areal capacity. Moreover, at the same areal capacity, electrodes with higher density exhibit higher cell energy density. Therefore, maximizing cell energy density requires simultaneous improvements in both areal capacity and electrode density. Previous studies have typically achieved high areal capacity or high electrode density individually, but rarely both simultaneously, limiting their applicability to practical, industrial-scale cells. In contrast, c-CNF-based electrodes show the potential to maximize cell energy density through the combination of high electrode density and high areal capacity, positioning the c-CNF binder as a key enabler for the development of energy-dense battery technologies.

This electrode design, featuring both high areal mass loading and high electrode density, demonstrates the potential to simplify battery architecture by reducing the number of unit cells and the amount of inactive components in pouch cells (details are described in Supplementary Fig. 41a, b). At an equivalent total cell capacity of 0.44 Ah, the PVDF-based pouch cell required three unit cells (each with an areal mass loading of 80 mg cm⁻²), whereas the c-CNF-based pouch cell achieved a comparable capacity using a single unit cell with an areal mass loading of 240 mg cm⁻² (Supplementary Fig. 41c). Here, each unit cell refers to a double-stacked electrode structure. As a result, the total cell thickness and weight of the c-CNF-based pouch cell were reduced by 51% and 27%, respectively (Fig. 7g, h), primarily due to the decreased use of inactive components such as current collectors and separators, thereby increasing the cell-level energy density. These structural advantages contribute to simplified cell assembly, higher material and volume efficiency, and ultimately serve as a foundation for the development of high-energy-density, cost-effective batteries.

Moreover, the c-CNF system enables high areal capacity electrodes, which translate into a lower practical cell-level cost. As shown in Fig. 7a, PVDF–based electrodes are practically limited to ~8 mAh cm⁻², whereas the c-CNF binder enables areal capacities up to 22 mAh cm⁻² (corresponding to 113 mg cm⁻²). Battery Performance and Cost (BatPaC) modeling^41,62 for a 72 Ah lithium-ion cell with a graphite||NCM811 configuration indicates that this increase in areal capacity reduces electrode manufacturing energy demand by 22.3% and lowers the overall cell-level cost by 5.8% (Fig. 7i). In addition to the areal capacity effect, we incorporated material and processing cost inputs into the BatPaC analysis based on the cost estimates summarized in Supplementary Table 7. Using these inputs, the estimated c-CNF binder cost is lower than that of PVDF, contributing to reduced fabrication cost. These results demonstrate high areal capacity electrodes enabled by the c-CNF binder as a viable pathway toward energy-dense and cost-competitive lithium-ion battery design.

Importantly, to further assess binder stability under practically relevant conditions, additional cycling tests were performed using graphite||NCM811 full cells. In these tests, full cells were assembled by pairing c-CNF–based NCM811 positive electrodes with commercial graphite negative electrodes. As shown in Supplementary Fig. 42, the c-CNF–based full cell exhibited stable cycling behavior, retaining approximately 88% of its initial capacity after 300 cycles, which is higher than that (~80%) of the PVDF–based control cell evaluated under identical conditions.

The c-CNF binder offers multifaceted advantages beyond electrochemical performance, improving several key metrics critical to advancing battery development (Fig. 7j and Supplementary Table 8). The c-CNF binder enables a unique combination of environmental sustainability (elimination of PFAS and NMP), reduced electrode tortuosity, enhanced $t_{{\mathrm{Li}}^{+}}$, high areal mass loading, and increased energy density, which is competitive with the conventional PVDF counterpart. These synergistic attributes position the c-CNF binder as a promising platform that seamlessly integrates sustainability with high-performance energy storage, offering a viable pathway toward environmentally friendly and scalable battery technologies.

Discussion

In summary, we presented a charge-engineered cellulose nanofibril (c-CNF) binder as a scalable and environmentally benign alternative to conventional PVDF systems that have relied on PFAS chemistry and hazardous solvents such as NMP. Leveraging its quaternary ammonium groups and nanofibrous structure, the c-CNF binder suppressed phase segregation, component agglomeration, and drying-induced cracking under high areal mass loading conditions, while concurrently ensuring efficient ionic and electronic transport. As a result, the c-CNF binder allowed the fabrication of high loading (113 mg cm⁻²) and high density (3.65 g cm⁻³) electrodes with an areal capacity of 22.5 mAh cm⁻² and a volumetric energy density of 1781.5 Wh L⁻¹, exhibiting competitive electrochemical performance relative to both PVDF-based counterparts and previously reported synthetic binders. In addition to its electrochemical benefits, the c-CNF binder enabled NMP-free slurry processing while remaining compatible with continuous roll-to-roll production, thereby reducing the environmental footprint of electrode manufacturing. By integrating precision molecular design with sustainable and scalable processing, the c-CNF offers a versatile binder platform applicable to a wide range of high-energy battery chemistries, including emerging multivalent systems.

Methods

Materials

Cellulose nanofibril suspension (CNF, high-fines slurry, solid content = 3 wt%) was purchased from the University of Maine. Single-walled carbon nanotubes (SWCNTs, HCNT4, Cabot, ≥99.7%) were used as received. Urea (≥99%), sodium hydroxide (NaOH, ≥99%), (3-chloro-2-hydroxypropyl)trimethylammonium chloride (60 wt% in H₂O, used as a cationic agent), ethylene glycol (EG, anhydrous, 99.8%), and N-methyl-2-pyrrolidone (NMP, 99.5%) were purchased from Sigma-Aldrich. Lithium bis(trifluoromethanesulfonyl)imide (LiTFSI, 99.0%) and polyvinylidene fluoride (PVDF, 280,000 g mol⁻¹) were purchased from Samchun and Solvay, respectively. Carbon black (Super C65) powders were supplied by Imerys. LiNi_0.8Co_0.1Mn_0.1O₂ (NCM811, LG Energy Solution) was used as the active material. 1 M lithium hexafluorophosphate (LiPF₆) in ethylene carbonate (EC)/ethyl methyl carbonate (EMC) (3:7, v/v), fluoroethylene carbonate (FEC, 99.9%), and vinylene carbonate (VC, 99.9%) was purchased from Enchem. All chemicals were used as received without further purification. Polyolefin (PE) separators (thickness = 20 μm) were provided by LG Energy Solution. Lithium metal foils (thickness = 100 μm, Honzo) were used as received, and handled in an argon-filled glovebox prior to cell assembly. Graphite negative electrode (mass loading: 12.2 mg cm⁻²; areal capacity: 4.28 mAh cm⁻²; porosity: 31%) was received from LG Energy Solution.

Preparation of c-CNF binder

Bare cellulose nanofibrils (b-CNF) refer to unmodified CNF. To obtain the anionic intermediate, a 1.0 wt% CNF suspension was dispersed in a NaOH/urea solution (7:12, w/w) diluted with deionized water, followed by cold storage at 4 °C for 12 h to induce partial deprotonation of hydroxyl groups and formation of alkoxide groups. The mixture was subsequently subjected to ultrasonication (20 kHz, 800 W) for 30 min^69–71. This state was defined as the anionic CNF (a-CNF). To introduce cationic functional groups, a quaternary ammonium-based cationic agent was added at predetermined molar ratios relative to the anhydroglucose units (AGUs), and the reaction was carried out under stirring at 70 °C for 8 h. The resulting charge-engineered CNF (c-CNF) was purified through dialysis against deionized water using cellulose dialysis tubing (Sigma-Aldrich) under continuous flow for 5 days until the pH stabilized at ~7.0. For anion exchange, 0.2 g of the synthesized c-CNF–Cl was immersed in 20 mL of 2 M aqueous LiTFSI solution and stirred at 70 °C for 72 h^56,72. The resulting c-CNF–TFSI was dialyzed to remove excess LiTFSI. The purified c-CNF–TFSI dispersion was then frozen and lyophilized under vacuum for 3 days using a freeze dryer equipped with a −120 °C condenser (FDCF-12012, Operon) to obtain the final binder.

Preparation of c-CNF binder-based positive electrodes

To prepare c-CNF-based positive electrodes, the freeze-dried c-CNF was dispersed in ethylene glycol (EG), used as a solvent, by ultrasonication. Carbon black (Super C65) additives were added and homogenized using a vortex mixer (10 min), followed by bath sonication (30 s). The slurry was then mixed with NCM811 at a typical composition ratio of 97:2:1 (NCM811/carbon black/c-CNF, w/w/w). Control electrodes were prepared using PVDF binder dissolved in NMP and mixed with NCM811 and carbon black at a ratio of 96:2:2. All slurries were single-side coated onto an Al foil (20 μm), using a doctor blade followed by roll pressing at 120 °C and vacuum drying at 120 °C for 12 h. For high-loading configurations, the c-CNF-based slurry was formulated with NCM811:SWCNT:c-CNF = 98.5:0.5:1 (w/w/w), in contrast to the electrode composition described above. The slurry was then cast, dried at 100 °C under vacuum for 12 h, and roll-pressed for densification. The resulting electrodes were fabricated at high areal mass loadings while maintaining a high electrode density of 3.65 g cm⁻³. The positive and negative electrodes were punched into circular disks with diameters of 10 mm and 14 mm, corresponding to an area of 0.79 cm² and 1.54 cm², respectively. Unless otherwise specified, the mass loading was 40 mg cm⁻², with different loadings indicated where applicable.

Structural and physicochemical characterization

High-resolution transmission electron microscopy (HR-TEM, ARM300, JEOL) was used to observe nanostructures. Fiber diameter distributions were quantified using ImageJ analysis. Surface and cross-sectional morphologies were analyzed using field-emission scanning electron microscopy (FE-SEM, SU-8100, Hitachi) at an accelerating voltage of 5–10 kV, and elemental analysis was performed using energy-dispersive X-ray spectroscopy (EDS, AZtecEnergy Ultim Max detector, Oxford Instruments) at 15 kV. FT–IR spectra (Varian 670-IR, Varian) were recorded from 400 to 4000 cm⁻¹. Elemental composition was measured with a CHNS analyzer (Flash2000, ThermoFisher). X-ray diffraction (XRD) was performed using a Rigaku SmartLab diffractometer with Cu Kα radiation (λ = 1.5406 Å). Patterns were collected in the 2θ range of 5–40° with a step size of 0.02° at a scan rate of 1.5° min⁻¹. Mechanical properties were characterized using a tensile tester (DA-01, Petrol LAB) at 10 mm min⁻¹ at 23 ± 1 °C. Specimens with an overall length of 70 mm, width of 10 mm, and thickness of 50 μm were tested, with a gauge length of 15 mm. Zeta potential was measured using a Zetasizer (Nano ZS, Malvern). Dispersions were prepared at a dilute concentration (10 ppm) in EG and NMP. Rheological measurements were performed using a rheometer (MCR 302, Anton Paar). Micro-computed tomography (SKYSCAN 1273, Bruker) was performed at 120 kV and 66 μA with a voxel size of 5.0 μm to visualize slurry microstructures. Scans were acquired over 360° with a step size of 0.3° and reconstructed using NRecon (Bruker). Laser scanning confocal microscopy (FV1000, Carl Zeiss) was used to assess surface uniformity and thickness variation. Field-emission electron probe microanalyser (EPMA, JXA-8530F, JEOL) mapping was conducted at 15 kV with a probe current of ~30 nA. Cycled NCM811 particles were cross-sectioned using a focused ion beam (FIB, Helios NanoLab 450, FEI). X-ray photoelectron spectroscopy (XPS, ESCALAB 250XI, Thermo Scientific) was performed using a monochromatic Al Kα source. Binding energies were calibrated to the C 1 s peak at 284.8 eV, and charge neutralization was applied. Depth profiling was conducted using Ar⁺ ion sputtering. The detection limit was ~0.1 at%, with a quantification accuracy of ±10%. Time-of-flight secondary ion mass spectrometry (ToF-SIMS, ToF-SIMS-5, ION TOF) measurements were performed using a Bi₃²⁺ primary ion beam operated at 30 keV with a current of 0.4 pA. Data were acquired in the negative ion mode using image mapping with an image size of 1024 × 1024 pixels. Charge compensation was applied using a flood gun during analysis. The measurements were conducted at 23 ± 1 °C. Cycled electrodes were harvested by disassembling the cells in an argon-filled glovebox, rinsed with dimethyl carbonate to remove residual electrolyte, dried, and transferred to the instrument using an airtight container to minimize air exposure. Electrical and interfacial resistivity were measured using an electrode analyzer (RM2610, Hioki). The adhesion strength of the positive electrodes was evaluated by peel test using a tensile tester (Petrol LAB DA-01) at a peel-off speed of 100 mm min⁻¹. ⁷Li magic-angle spinning (MAS) nuclear magnetic resonance (NMR) spectroscopy was employed to probe the local Li⁺ environment in the electrodes. Spectra were acquired on a 600 MHz FT–NMR spectrometer (VNMRS 600, Agilent) using a 1.6 mm HXY Fast MAS T3 probe at a spinning rate of 20 kHz. Chemical shifts were referenced to an external 1.0 M LiCl aqueous solution (0 ppm). Through-thickness Raman spectroscopy (NRS-3100, JASCO) was employed to assess the degree of delithiation and post-cycling structural degradation of the electrodes. The redox state of NCM811 was analyzed using the E_g/A_1g vibrational mode intensity ratio. All measurements were conducted at 23 ± 1 °C.

Electrochemical measurements

The electrochemical performance of the positive electrodes was evaluated using 2032-type coin cells (assembled in stainless steel cases with springs) and pouch-type (electrode area = 2 × 2 cm²) cells, which consisted of either a PVDF or c-CNF electrode, a polyethylene separator (20 μm, Toray-Tonen), and Li-metal (100 μm). The cell assembly was conducted in an argon-filled glovebox. Two types of pouch cells were fabricated. (1) Standard pouch cells (used for Fig. 6g and related measurements): These cells were constructed in a single-layer configuration using NCM811 positive electrodes with an areal mass loading of 40 mg cm⁻², paired with a Li-metal negative electrode (100 μm) on a Cu current collector. The nominal capacity of these cells was approximately 0.033 Ah. (2) Areal-capacity-matched pouch cells (used for Fig. 7g, h and Supplementary Fig. 41): Pouch cells were assembled to achieve a comparable total cell capacity (~0.44 Ah) by varying the number of unit cells. PVDF-based electrodes were assembled using three unit cells, whereas c-CNF-based electrodes were assembled using a single unit cell with an increased electrode loading. Detailed electrode loadings and configurations are provided in the Supplementary Information. A liquid electrolyte composed of 1 M LiPF₆ in EC/EMC (3:7, v/v) with 10 wt% FEC and 2 wt% VC additives was used, with an electrolyte volume of 100 μL per coin cell and 150 μL per pouch cell (corresponding to an electrolyte-to-capacity (E/C) ratio of 4.5 mL Ah⁻¹ for the standard pouch cells). All electrochemical measurements were conducted at 23 ± 1 °C. Linear sweep voltammetry (LSV) measurements were performed at a sweep rate of 1 mV s⁻¹. The applied potential range was from the open-circuit voltage (OCV) to 5.5 V. The electrochemical performance was assessed using a battery cycler (PNE Solution) under various charge/discharge conditions. Galvanostatic charge–discharge tests were performed within a voltage range of 3.0–4.3 V (vs. Li⁺/Li) at various C-rates, with all currents expressed as specific currents based on the mass of active material (1 C = 200 mA g⁻¹). No pre-activation of the negative electrode was performed prior to full cell assembly. Coulombic efficiency (CE) was calculated as the ratio of the discharge capacity to the charge capacity of the preceding charge cycle. Additionally, the electrochemical impedance spectroscopy (EIS) and galvanostatic intermittent titration technique (GITT) measurements were performed using a potentiostat/galvanostat (VSP Classic, Bio-Logic) in a symmetric cell (electrode||electrode) configuration to analyze the kinetic and interfacial properties of the cells. EIS measurements were performed in potentiostatic mode over a frequency range of 1 MHz to 0.01 Hz with an AC amplitude of 14.1 mV. The spectra were recorded with 10 points per decade. Prior to the measurements, the cells were rested at open-circuit voltage (OCV) for 24 h to ensure quasi-stationary conditions. The Li⁺ transference number ($t_{{\mathrm{Li}}^{+}}$) was determined using potentiostatic polarization combined with EIS. A constant potential difference (ΔV) was applied to a Li/Li symmetric cell, and the resulting current response was monitored until a steady state was reached. Impedance spectra were obtained before and after polarization to extract the interfacial resistances. The $t_{{\mathrm{Li}}^{+}}$ value was calculated according to the Bruce–Vincent–Evans equation⁷³:

$$t_{Li}=\frac{I_s\left(\mathrm{\Delta }V-I_0{R}_0\right)}{I_0\left(\mathrm{\Delta }V-I_s{R}_s\right)}$$

where I₀ and I_s represent the initial and steady-state currents, respectively, and R₀ and R_s correspond to the interfacial resistances measured before and after polarization, respectively. The corresponding parameters used for the calculation are summarized in Supplementary Table 2. GITT measurements were performed using repeated current pulses at 0.5 C (4 mA cm⁻²) with a 60 min relaxation period between pulses. The data acquisition was primarily event-driven, with recording triggered by potential changes (10 mV) in addition to a time interval of 120 s. The GITT analysis was used to evaluate polarization/internal resistance, and diffusion coefficients were not calculated; therefore, validation of diffusion coefficient calculations based on potential–concentration linearity is not applicable. The pouch-type cells, comprising a Li-metal negative electrode and an NCM811 positive electrode (areal capacity: 8.0 mAh cm⁻²; negative/positive capacity ratio (N/P = 2.5)), were assembled and sealed in aluminum pouch films in a dry room (dew point: −60 °C). Unless otherwise specified, electrochemical data are presented from a representative cell selected based on its consistency with the typical behavior observed across independently tested cells (typically n ≥ 3). The specific energy and energy density of the pouch cell were determined using the experimentally measured cell mass and volume. The total cell weight included all components, including positive and negative electrodes, separator, electrolyte, and current collectors. The detailed methodology for these calculations is provided in Supplementary Tables 4 and 5.

Density functional theory calculations

Density functional theory (DFT) calculations were performed using the VASP simulation package to evaluate the interaction energy between the binders in a vacuum environment (Supplementary Data 1, 2, 10–12). The generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional was employed^74–76. A Gamma-point calculation was conducted using the projector-augmented-wave (PAW) method. The plane-wave basis set had an energy cutoff of 450 eV. The electronic self-consistent field (SCF) convergence criterion, EDIFF, was set to 1 × 10⁻⁶ eV, and the sigma parameter was set to 0.05⁷⁷.

Molecular dynamics simulations

All-atom molecular dynamics (MD) simulations were performed using GROMACS^78,79. Force field parameters for PVDF, CNF, and c-CNF were generated using the LigParGen^80–82 platform based on optimized potentials for liquid simulations all-atom (OPLS-AA) force field with 1.14* CM1A charge model. Parameters for the ions were taken from Canongia Lopes–Pádua force field (CL&P FF), and their charges were scaled down to 0.8⁸³. Parameters for the Al₂O₃ crystal slab were taken from ClayFF⁸⁴ force field and the Materials Project mp-7048^85,86. A cutoff distance of 1 nm was applied for short-range Lennard–Jones and Coulomb interactions, whereas long-range electrostatics were treated using the particle mesh Ewald (PME) approach⁸⁷. The LINCS algorithm⁸⁸ was employed to constrain bond lengths. Simulations were performed in the NVT ensemble at 333 K with a 2 fs integration time step, controlled by a Nosé–Hoover thermostat^89,90. The interaction energy between the Al₂O₃ surface and the binder was calculated based on the energy difference between the isolated state and the stabilized state when present together, given by $\mathrm{\Delta }E=E_{\mathrm{t}\mathrm{ot}}-E_{{\mathrm{A}\mathrm{l}}_2{\mathrm{O}}_3}-E_{\mathrm{b}\mathrm{inder}}$, where $E_{\mathrm{t}\mathrm{ot}}$ is the total energy of the Al₂O₃–binder system, $E_{{\mathrm{A}\mathrm{l}}_2{\mathrm{O}}_3}$ is the energy of Al₂O₃, and $E_{\mathrm{b}\mathrm{inder}}$ is the energy of the binder in a vacuum state (Supplementary Data 3–9). To account for various interacting configurations between the Al₂O₃ surface and the binder, the energy of 40 different interacting structures was calculated and averaged. Each system underwent a 10 ns NVT simulation, with the last 2 ns used for energy calculations. For single-component vacuum simulations, a 200 ns NVT simulation was performed, and the last 100 ns was used for energy calculation. For the analysis of the binder–PF₆ radial distribution function (RDF, g(r)) and coordination number (CN), systems consisting of a 6 nm cubic box filled with 1 M LiPF₆ in EC/EMC (1:1, v/v) were prepared (Supplementary Data 13–16). For simplicity, the MD simulations were performed using 1 M LiPF₆ in EC/EMC (1:1, v/v) electrolyte, which differs from the experimental composition (3:7, v/v). To each system, 5, 10, 20, and 40 chains of PVDF or c-CNF (both set to a comparable molecular weight) were added. Following equilibration, each system underwent a 1 ns NVT simulation, a 10 ns NPT simulation, and a final 30 ns NVT simulation. The last 20 ns were used for analysis, and the results for the systems containing 40 chains are presented.

Author contributions

S.-W.K., N.-Y.K., and S.-Y.L. designed this work. S.-W.K. and N.-Y.K. performed the experimental characterization and electrochemical tests. Y.J.K., S.-S.C., S.-C.N., C.C., Y.M.L., J.-I.C., and J.-H.K. assisted with selected experiments and data analysis. A.P. and J.H. performed the theoretical calculations and modeling simulations. C.B. conducted the SAICAS analysis. W.B.L. and S.-Y.L. supervised the overall project. All authors contributed to finalizing the manuscript.

Peer review

Peer review information

Nature Communications thanks An-Min Cao and the other, anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the Basic Science Research Program (RS-2023-00261543, RS-2024-00344021, RS-2024-00455177, and RS-2025-25441257) through the National Research Foundation of Korea (NRF) granted by the Korean Government (MSIT) and supported by POSCO Holdings. This work was also supported by Korea Institute for Advancement of Technology (KIAT) grant funded by the Korea Government (MOTIE) (RS-2024-00420590, HRD Program for Industrial Innovation).

Data availability

The data supporting the findings of this study are available in the paper, the Supplementary Information, and the Supplementary Data. The optimized atomic coordinates for DFT calculations and the initial and final configurations for MD simulations are provided in Supplementary Data 1–16. Additional data are available from the corresponding author upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-73909-0.